Files
ai_platform/lib/parse-checklist-from-minutes.js

355 lines
9.8 KiB
JavaScript

/**
* 회의록 generated_minutes(Markdown)에서 체크리스트·액션 섹션 추출
* LLM 출력은 ##/### 혼용, 번호 접두, 굵게 표기 등이 섞이므로 규칙을 넓게 둔다.
*/
const CHECKLIST_HEADINGS = [
/^##\s+후속\s*확인\s*체크리스트\s*$/i,
/^##\s+체크리스트\s*\([^)]*\)\s*$/i,
/^##\s+체크리스트\s*$/i,
/^##\s+후속\s*확인\s*$/i,
];
const ACTION_HEADINGS = [
/^##\s+Action\s+Items\s*$/i,
/^##\s+Action\s+Item\s*$/i,
/^##\s+액션\s*아이템\s*$/i,
/^##\s+액션\s*$/i,
];
function stripMd(s) {
if (!s) return "";
return s
.replace(/\*\*([^*]+)\*\*/g, "$1")
.replace(/__([^_]+)__/g, "$1")
.trim();
}
/**
* 제목 줄에서 비교용 텍스트 (굵게·번호 접두 제거)
* @param {string} line
*/
function normalizeHeadingText(line) {
return line
.trim()
.replace(/^#{1,6}\s+/, "")
.replace(/^\d+[.)]\s*/, "")
.replace(/\*\*/g, "")
.replace(/\s+/g, " ")
.trim()
.toLowerCase();
}
/**
* @param {string} trimmed — 한 줄 (마크다운 제목 가능)
*/
function isChecklistHeadingLine(trimmed) {
const t = normalizeHeadingText(trimmed);
if (!t) return false;
if (/액션\s*아이템/.test(t) && /체크리스트/.test(t)) return true;
if (/^액션\s*아이템\s*$/.test(t) || /^action\s+items\s*$/.test(t)) return false;
if (t.includes("체크리스트")) return true;
if (t.includes("후속") && (t.includes("확인") || t.includes("체크"))) return true;
if (t.includes("follow") && t.includes("check")) return true;
if (t.includes("follow-up") || t.includes("follow up")) return true;
return false;
}
/**
* @param {string} trimmed
*/
function isActionHeadingLine(trimmed) {
const t = normalizeHeadingText(trimmed);
if (!t) return false;
if (/^액션\s*아이템\s*$/.test(t) || /^action\s+items\s*$/i.test(t) || /^action\s+item\s*$/i.test(t)) return true;
if (t.includes("액션") && (t.includes("아이템") || t.includes("항목"))) return true;
if (t.includes("action") && (t.includes("item") || t.includes("items"))) return true;
return false;
}
/**
* @param {string} text
* @param {(trimmedLine: string) => boolean} predicate
* @returns {string|null}
*/
function extractSectionByHeadingPredicate(text, predicate) {
const lines = text.split(/\r?\n/);
for (let i = 0; i < lines.length; i++) {
const trimmed = lines[i].trim();
const hm = trimmed.match(/^(#{1,6})(\s+.+)$/);
if (!hm) continue;
const level = hm[1].length;
if (!predicate(trimmed)) continue;
const body = [];
for (let j = i + 1; j < lines.length; j++) {
const tj = lines[j].trim();
const nextHm = tj.match(/^(#{1,6})\s+/);
if (nextHm) {
const nextLevel = nextHm[1].length;
if (nextLevel <= level) break;
}
body.push(lines[j]);
}
const joined = body.join("\n").trim();
if (joined.length) return joined;
}
return null;
}
/**
* 다음 ## 제목 전까지 본문 추출 (레거시: ## 만 구분)
* @param {string} text
* @param {RegExp[]} headingMatchers
* @returns {string|null}
*/
function extractSectionAfterHeading(text, headingMatchers) {
const lines = text.split(/\r?\n/);
for (let i = 0; i < lines.length; i++) {
const trimmed = lines[i].trim();
for (const re of headingMatchers) {
if (re.test(trimmed)) {
const body = [];
for (let j = i + 1; j < lines.length; j++) {
if (/^##\s+/.test(lines[j].trim())) break;
body.push(lines[j]);
}
const joined = body.join("\n").trim();
return joined.length ? joined : null;
}
}
}
return null;
}
/**
* @typedef {{ title: string, detail: string, assignee: string|null, due_note: string|null, completed: boolean }} ParsedItem
*/
/**
* @param {string} body
* @returns {ParsedItem[]}
*/
function parseBulletItems(body) {
if (!body || !body.trim()) return [];
const lines = body.split(/\r?\n/);
/** @type {ParsedItem[]} */
const items = [];
/** @type {ParsedItem|null} */
let cur = null;
const flush = () => {
if (cur) {
items.push(cur);
cur = null;
}
};
for (const raw of lines) {
const t = raw.trim();
if (!t) continue;
let m = t.match(/^\s*[-*•]\s+\[([ xX✓])\]\s*(.+)$/);
if (m) {
flush();
items.push({
title: stripMd(m[2].trim()),
detail: "",
assignee: null,
due_note: null,
completed: /[xX✓]/.test(m[1]),
});
continue;
}
m = t.match(/^\s*\[\s*([ xX✓])\s*\]\s+(.+)$/);
if (m) {
flush();
items.push({
title: stripMd(m[2].trim()),
detail: "",
assignee: null,
due_note: null,
completed: /[xX✓]/.test(m[1]),
});
continue;
}
m = t.match(/^\s*[☐☑✓✔]\s*(.+)$/);
if (m) {
flush();
const done = /^[☑✓✔]/.test(t.trim());
items.push({
title: stripMd(m[1].trim()),
detail: "",
assignee: null,
due_note: null,
completed: done,
});
continue;
}
m = t.match(/^\s*[-*•]\s+(.+)$/);
if (m) {
flush();
cur = { title: stripMd(m[1].trim()), detail: "", assignee: null, due_note: null, completed: false };
continue;
}
m = t.match(/^\s*\d+\.\s+(.+)$/);
if (m) {
flush();
cur = { title: stripMd(m[1].trim()), detail: "", assignee: null, due_note: null, completed: false };
continue;
}
if (cur) {
cur.detail += (cur.detail ? "\n" : "") + t;
} else if (items.length) {
const last = items[items.length - 1];
last.detail += (last.detail ? "\n" : "") + t;
}
}
flush();
return items.filter((x) => x.title.length > 0);
}
/**
* 액션 아이템 번호 목록 블록 (담당/기한/할 일)
* @param {string} body
* @returns {ParsedItem[]}
*/
function refineActionLines(detailText) {
let assignee = null;
let due_note = null;
const rest = [];
for (const line of (detailText || "").split(/\r?\n/)) {
const r = line.trim();
if (!r) continue;
const t = r.replace(/^\*\s+/, "").trim();
if (/담당\s*:/i.test(t)) {
const m = t.match(/담당\s*:\s*(.+)$/i);
if (m) assignee = stripMd(m[1].replace(/\*\*/g, "").replace(/^\*+|\*+$/g, "").trim());
continue;
}
if (/기한\s*:/i.test(t)) {
const m = t.match(/기한\s*:\s*(.+)$/i);
if (m) due_note = stripMd(m[1].replace(/\*\*/g, "").replace(/^\*+|\*+$/g, "").trim());
continue;
}
if (/할\s*일\s*:/i.test(t)) {
const m = t.match(/할\s*일\s*:\s*(.+)$/i);
if (m) rest.push(stripMd(m[1]));
continue;
}
rest.push(r);
}
return { assignee, due_note, detail: rest.join("\n").trim() };
}
function parseNumberedActionBlocks(body) {
if (!body || !body.trim()) return [];
const lines = body.split(/\r?\n/);
/** @type {ParsedItem[]} */
const out = [];
let i = 0;
while (i < lines.length) {
const line = lines[i].trim();
/** `1. 제목` 또는 `1) 제목` */
const nm = line.match(/^(\d+)[.)]\s+(.+)$/);
if (nm) {
const title = stripMd(nm[2].trim());
const rest = [];
i++;
while (i < lines.length) {
const lt = lines[i].trim();
if (/^\d+[.)]\s+/.test(lt)) break;
if (lt) rest.push(lines[i]);
i++;
}
const rawDetail = rest.join("\n").trim();
const refined = refineActionLines(rawDetail);
out.push({
title,
detail: refined.detail,
assignee: refined.assignee,
due_note: refined.due_note,
completed: false,
});
continue;
}
i++;
}
return out.filter((x) => x.title.length > 0);
}
/**
* 업무 체크리스트 자동 동기화용: 규칙 기반으로 액션(번호 목록) → 체크리스트 순으로 항목 수집(제목 기준 중복 제거)
* @param {string} markdown
* @returns {ParsedItem[]}
*/
function parseAllRuleBasedWorkItems(markdown) {
const text = (markdown || "").trim();
if (!text) return [];
const actions = parseItemsFromMinutes(text, "actions");
const checklist = parseItemsFromMinutes(text, "checklist");
const seen = new Set();
/** @type {ParsedItem[]} */
const out = [];
for (const it of [...actions, ...checklist]) {
const t = (it.title || "").trim().toLowerCase();
if (!t) continue;
if (seen.has(t)) continue;
seen.add(t);
out.push(it);
}
return out;
}
/**
* 체크리스트 섹션 본문 찾기: 유연 매칭 → 레거시 정규식
* @param {string} text
* @returns {string|null}
*/
function extractChecklistSectionBody(text) {
const flexible = extractSectionByHeadingPredicate(text, isChecklistHeadingLine);
if (flexible) return flexible;
return extractSectionAfterHeading(text, CHECKLIST_HEADINGS);
}
/**
* @param {string} text
* @returns {string|null}
*/
function extractActionSectionBody(text) {
const flexible = extractSectionByHeadingPredicate(text, isActionHeadingLine);
if (flexible) return flexible;
return extractSectionAfterHeading(text, ACTION_HEADINGS);
}
/**
* @param {string} generatedMinutes
* @param {'checklist'|'actions'} mode
* @returns {ParsedItem[]}
*/
function parseItemsFromMinutes(generatedMinutes, mode = "checklist") {
const text = (generatedMinutes || "").trim();
if (!text) return [];
if (mode === "actions") {
const section = extractActionSectionBody(text);
if (!section) return [];
const numbered = parseNumberedActionBlocks(section);
if (numbered.length) return numbered;
return parseBulletItems(section);
}
const section = extractChecklistSectionBody(text);
if (!section) return [];
return parseBulletItems(section);
}
module.exports = {
extractSectionAfterHeading,
extractSectionByHeadingPredicate,
extractChecklistSectionBody,
extractActionSectionBody,
parseBulletItems,
parseNumberedActionBlocks,
parseItemsFromMinutes,
parseAllRuleBasedWorkItems,
CHECKLIST_HEADINGS,
ACTION_HEADINGS,
isChecklistHeadingLine,
isActionHeadingLine,
};